The workshop on Scholarly Document Processing (SDP) started in 2020 to accelerate research, inform policy and educate the public on natural language processing for scientific text. The fourth iteration of the workshop, SDP24 was held at the 62nd Annual Meeting of the Association for Computational Linguistics (ACL24) as a hybrid event. The SDP workshop saw a great increase in interest, with 57 submissions, of which 28 were accepted. The program consisted of a research track, four invited talks and two shared tasks: 1) DAGPap24: Detecting automatically generated scientific papers and 2) Context24: Multimodal Evidence and Grounding Context Identification for Scientific Claims. The program was geared towards NLP, information extraction, information retrieval, and data mining for scholarly documents, with an emphasis on identifying and providing solutions to open challenges.

We propose a method to teach multiple large language models (LLM) to collaborate by interleaving their generations at the token level. We model the decision of which LLM generates the next token as a latent variable. By optimizing the marginal likelihood of a training set under our latent variable model, the base LLM automatically learns when to generate itself and when to call on one of the “assistant” language models to generate, all without direct supervision. Token-level collaboration during decoding allows for a fusion of each model’s expertise in a manner tailored to the specific task at hand. Our collaborative decoding is especially useful in cross-domain settings where a generalist base LLM learns to invoke domain expert models. On instruction-following, domain-specific QA, and reasoning tasks, we show that the performance of the joint system exceeds that of the individual models. Through qualitative analysis, we show models trained with our method exhibit several interesting collaboration patterns, e.g., template-filling, by visualizing the learned latent decisions.

Information about pretraining corpora used to train the current best-performing language models is seldom discussed: commercial models rarely detail their data, and even open models are often released without accompanying training data or recipes to reproduce them. As a result, it is challenging to conduct and advance scientific research on language modeling, such as understanding how training data impacts model capabilities and limitations. To facilitate scientific research on language model pretraining, we curate and release Dolma, a three-trillion-token English corpus, built from a diverse mixture of web content, scientific papers, code, public-domain books, social media, and encyclopedic materials. We extensively document Dolma, including its design principles, details about its construction, and a summary of its contents. We present analyses and experimental results on intermediate states of Dolma to share what we have learned about important data curation practices. Finally, we open-source our data curation toolkit to enable reproduction of our work as well as support further research in large-scale data curation.

Recent work has shown that infusing layout features into language models (LMs) improves processing of visually-rich documents such as scientific papers. Layout-infused LMs are often evaluated on documents with familiar layout features (e.g., papers from the same publisher), but in practice models encounter documents with unfamiliar distributions of layout features, such as new combinations of text sizes and styles, or new spatial configurations of textual elements. In this work we test whether layout-infused LMs are robust to layout distribution shifts. As a case study we use the task of scientific document structure recovery, segmenting a scientific paper into its structural categories (e.g., “title”, “caption”, “reference”). To emulate distribution shifts that occur in practice we re-partition the GROTOAP2 dataset. We find that under layout distribution shifts model performance degrades by up to 20 F1. Simple training strategies, such as increasing training diversity, can reduce this degradation by over 35% relative F1; however, models fail to reach in-distribution performance in any tested out-of-distribution conditions. This work highlights the need to consider layout distribution shifts during model evaluation, and presents a methodology for conducting such evaluations.

Despite growing interest in applying natural language processing (NLP) and computer vision (CV) models to the scholarly domain, scientific documents remain challenging to work with. They’re often in difficult-to-use PDF formats, and the ecosystem of models to process them is fragmented and incomplete. We introduce PaperMage, an open-source Python toolkit for analyzing and processing visually-rich, structured scientific documents. PaperMage offers clean and intuitive abstractions for seamlessly representing and manipulating both textual and visual document elements. PaperMage achieves this by integrating disparate state-of-the-art NLP and CV models into a unified framework, and provides turn-key recipes for common scientific document processing use-cases. PaperMage has powered multiple research prototypes of AI applications over scientific documents, along with Semantic Scholar’s large-scale production system for processing millions of PDFs. GitHub: https://github.com/allenai/papermage

Layout detection is an essential step for accurately extracting structured contents from historical documents. The intricate and varied layouts present in these document images make it expensive to label the numerous layout regions that can be densely arranged on each page. Current active learning methods typically rank and label samples at the image level, where the annotation budget is not optimally spent due to the overexposure of common objects per image. Inspired by recent progress in semi-supervised learning and self-training, we propose OLALA, an Object-Level Active Learning framework for efficient document layout Annotation. OLALA aims to optimize the annotation process by selectively annotating only the most ambiguous regions within an image, while using automatically generated labels for the rest. Central to OLALA is a perturbation-based scoring function that determines which objects require manual annotation. Extensive experiments show that OLALA can significantly boost model performance and improve annotation efficiency, facilitating the extraction of masses of structured text for downstream NLP applications.

Accurately extracting structured content from PDFs is a critical first step for NLP over scientific papers. Recent work has improved extraction accuracy by incorporating elementary layout information, for example, each token’s 2D position on the page, into language model pretraining. We introduce new methods that explicitly model VIsual LAyout (VILA) groups, that is, text lines or text blocks, to further improve performance. In our I-VILA approach, we show that simply inserting special tokens denoting layout group boundaries into model inputs can lead to a 1.9% Macro F1 improvement in token classification. In the H-VILA approach, we show that hierarchical encoding of layout-groups can result in up to 47% inference time reduction with less than 0.8% Macro F1 loss. Unlike prior layout-aware approaches, our methods do not require expensive additional pretraining, only fine-tuning, which we show can reduce training cost by up to 95%. Experiments are conducted on a newly curated evaluation suite, S2-VLUE, that unifies existing automatically labeled datasets and includes a new dataset of manual annotations covering diverse papers from 19 scientific disciplines. Pre-trained weights, benchmark datasets, and source code are available at https://github.com/allenai/VILA.

Abstractive summarization systems today produce fluent and relevant output, but often “hallucinate” statements not supported by the source text. We analyze the connection between hallucinations and training data, and find evidence that models hallucinate because they train on target summaries that are unsupported by the source. Based on our findings, we present PINOCCHIO, a new decoding method that improves the consistency of a transformer-based abstractive summarizer by constraining beam search to avoid hallucinations. Given the model states and outputs at a given step, PINOCCHIO detects likely model hallucinations based on various measures of attribution to the source text. PINOCCHIO backtracks to find more consistent output, and can opt to produce no summary at all when no consistent generation can be found. In experiments, we find that PINOCCHIO improves the consistency of generation by an average of 67% on two abstractive summarization datasets, without hurting recall.

Adobe’s Portable Document Format (PDF) is a popular way of distributing view-only documents with a rich visual markup. This presents a challenge to NLP practitioners who wish to use the information contained within PDF documents for training models or data analysis, because annotating these documents is difficult. In this paper, we present PDF Annotation with Labels and Structure (PAWLS), a new annotation tool designed specifically for the PDF document format. PAWLS is particularly suited for mixed-mode annotation and scenarios in which annotators require extended context to annotate accurately. PAWLS supports span-based textual annotation, N-ary relations and freeform, non-textual bounding boxes, all of which can be exported in convenient formats for training multi-modal machine learning models. A read-only PAWLS server is available at https://pawls.apps.allenai.org/, and the source code is available at https://github.com/allenai/pawls.